IAMCCS CineInfo V2V

The info node at the head of a V2V segment pipeline

Video-to-video with LTX-2 is a segment-by-segment affair: each pass takes source footage, conditioning, and audio, and re-renders a few seconds at a time with continuity. Somewhere you need a node that knows what this segment's generation parameters, source material, and prompt actually are. That's this node's entire job. It's an information faucet - you feed it the pipeline's cine_linx (the contract object the IAMCCS Cine/SuperNode system threads through the graph) and it unpacks everything downstream stages need.

What comes out

The output list is huge, but they group into a handful of jobs:

• Generation parameters - fps, generation_width, generation_height, continuation_loops. These feed your sampler and latent setup directly.
• The prompt - global_prompt and negative_prompt come from the linx contract, so your planner's decisions land exactly where the conditioning node can read them.
• Source material - source_image, source_images, source_video_images, source_audio, and source_video_info (a VHS_VIDEOINFO, so this expects VideoHelperSuite to be around). Plus current_source_images and current_segment_audio, which are the sliced-down versions for this segment.
• Segment state - segment_index_out, planner_report, current_audio_report, and a general report.
• The heavy one - audio_vae. Set load_audio_vae to true and the node loads the LTX audio VAE and hands it to you, saving you a separate loader node in audio-driven V2V.

The inputs you'll touch

Only a few: cine_linx is required and non-negotiable - without it the node has nothing to read. segment_index lets you probe a specific segment rather than the current one. load_audio_vae is the useful toggle. source_images and source_audio are optional overrides when the contract doesn't already carry them.

Honest framing

This is a glue node, not a headline feature. You reach for it when you're building or debugging an IAMCCS V2V workflow and you're tired of hunting for where fps or the segment's audio actually comes from - you drop CineInfo after the planner, and suddenly every value is inspectable. It's also a great debugging surface: the report and current_audio_report strings tell you what the contract believes about the current segment, which is where most V2V problems actually hide.

Install is the pack routine (Manager: "IAMCCS", or clone IAMCCS/IAMCCS-nodes into custom_nodes, restart). Watch the dependency chain: the VHS_VIDEOINFO output and audio VAE handling assume VideoHelperSuite and a working LTX audio-VAE model are present. On its own it's useless - but in a Cine V2V graph it's the difference between guessing what the pipeline thinks and knowing.